As the world becomes more connected, the reliance on precise positioning and environmental awareness is growing. A prime example is ride-hailing applications on our smartphones, in which a relatively accurate position is required for both the person requesting the ride and the vehicle collecting them. Due to the urbanised environments these services often operate in, this need for precision poses a challenge. The autonomous vehicles currently being developed face similar challenges to ride-hailing apps, in that they need accurate positioning to function. In this case, though, a loss of precision changes from an inconvenience to a critical – and potentially lethal – flaw. Moreover, as new regulations come into force for systems which require position and situational awareness, accurate positioning testing and assurance becomes even more critical.
Absolute and relative position
When we talk about positioning, we are referring to one or both of two different types: absolute and relative. GNSS is still the only globally available absolute positioning source and is utilised in functions such as eCall and Intelligent Speed Assistance. However, as GNSS has limitations, it is generally partnered with a robust inertial measurement unit (IMU). Other sensors such as cameras, lidar and radar, alongside HD-maps, are then used to complement the more traditional GNSS/inertial setup. The IMU and additional sensors only operate relative to the position provided by the absolute source.
These integrated systems need to be properly tested and simultaneously validated to guarantee the implementation of a robust and precise positioning system.
Accuracy, integrity and availability
Safety-critical positioning systems specify their performance in terms of accuracy, integrity and availability. Accuracy, in this definition, is a measure of nominal performance, and defines the maximum allowable error for operation. Integrity is a measure of the system limits and the probability of system failure. Integrity is generally defined as probability of failure per hour of operation. Availability is a measure of how consistently the system is operating within the maximum permissible error range. A system has 100% availability if it is constantly operating within these limits.
When operating in Autonomous Mode, precise position must always be available. Some sensors and systems are better at providing lateral information, such as cameras recognising lane lines, and other types of sensors - such as wheel odometry - can provide longitudinal information. In the world of autonomous vehicles, the GNSS receiver is just one subsystem in a complex system of sensors, actuators, software applications and algorithms. At present, there is no single technology that can meet the requirements for safe operation in all weather, road and traffic scenarios. Therefore, several localisation systems must be fused together to achieve the intended system requirements.
In 2019 we saw the introduction of eCall which automatically calls the emergency services and reports the vehicle position. As from July 2022, Intelligent Speed Assistance (ISA) is mandatory in new cars in the EU, but its impact is not limited to that region of the world.
Forward-facing cameras, Global Navigation Satellite System (GNSS) for positioning the vehicle on the road, an electronic horizon, and map data that includes all speed limits — whether visibly marked with signs or not — are the minimum requirement for ISA systems today.
Why do we need realism in simulation?
Any system integrator, developer or user will want to be assured that the vehicle’s sensor fusion localisation system will function effectively within its expected environment. Systems must be designed and tested to ensure that degradation of any sensor can be detected - while maintaining safe operation.
There currently is a great deal of momentum driving radar, lidar and camera development and testing. Realistic scenarios created to validate the system generally include deteriorating weather conditions, environment changes, other vehicles and road users, and many more real-world variables to test how gracefully sensor systems degrade. Due to the above “new” mandatory features, more rigorous GNSS scenario testing, as opposed to the more traditional functional testing is now required.
There is a tendency to treat GNSS differently to these other sensors, and fail to expose it to the necessary sources of real-world degradation when testing. Similarly to radar, lidar and cameras, GNSS signals can become unavailable (in a tunnel or an indoor carpark, for instance), and receivers can also be compromised by anything from tall buildings to deliberate jamming and spoofing by a threat source. Therefore, GNSS must be tested with the same level of attention as any other sensor on the autonomous vehicle. Each element of the vehicle positioning, navigation and timing (PNT) system must be tested to absolute regress; and to do this properly developers must approach this using an integrated and holistic testing solution.
Moving in the right direction with co-simulation
Testing the fusion of multiple positioning sources is a new requirement for autonomous vehicle developers. A co-simulation platform must simultaneously simulate all the sources of absolute and relative position, as well as a realistic environment. By using a positioning co-simulation platform, a vehicle developer can test real-world scenarios to ensure safe and reliable design of the vehicle navigation systems. As well as the normal environment, co-simulation should be able to simulate many different types of interference and disruption, weather conditions and potential threats which are fed to all the subsystems concurrently.
If you’re interested to learn more about positioning, navigation and timing for connected and autonomous vehicles, watch our on-demand webinar, where Raphael is presenting.
This blog was originally published in August 2019 and republished with updates.